An update on a previously published study that looks at forecasting volatility using the S&P 500 Volatility Index (VIX) and the VIX Volatility Index (VVIX)
About a month ago, we published an article titled “Volatility on the Horizon?” which is based on a study called “Forecasting a Volatility Tsunami” in which the author examines forecasting a spike in volatility using the CBOE S&P 500 Volatility Index (VIX). The paper found that spikes in volatility are typically preceded by extended periods when the VIX is subdued. The author attempted to improve reliability by including the CBOE VIX Volatility Index (VVIX), which measures the expected volatility of the VIX itself, i.e., the volatility of volatility. They found that almost every large spike in volatility had been preceded by the 20-day standard deviation of the VIX and VVIX simultaneously falling to extreme low levels, which they defined as below the 15th percentile.
The original paper and our updated study looked only at the increase in volatility around these periods. But after we published the article we received requests to look at how the market behaved during these periods. So, today we have updated our study to include the most recent period of low VIX & VVIX readings and looked at the drawdown the S&P has experienced around these periods.
The chart below shows the daily closing values of the VIX, and, in red, the times when VIX and VVIX’s 20-day standard deviation were below the 15th percentile. As you can see, there have been multiple occasions when both indexes breached the 15th percentile threshold since the end of the original study period in 2016. In some instances, there were relatively moderate volatility spikes, while other times, like early 2018, the VIX rose by more than 100%. You will also notice that there is a cluster of red at the end of the graph as both indexes’ 20-day standard deviation fell below the 15th percentile in the final week of August through the first week of September.
To get a clearer picture of how the VIX behaves following periods when both index’s 20-day standard deviation falls below the 15th percentile, we looked out at the maximum increase and decrease in the VIX over the next 30 days. Because the sub-15 periods are often clusters of days, we looked 30 days forward from both the beginning of the cluster and the end of the cluster (only in two instances were there single-day periods.) The results are summarized in the table below and as you can see there was a noticeable bias towards the VIX rising over the next 30 days with an average increase of 28% from the beginning of the sub-15 period reading and a 46% average increase from the end of the period.
For this study we also looked at how the S&P has behaved around these periods. Volatility spikes are typically accompanied by declines in the equity market. To get an idea of the magnitude of these drawdowns we took a similar approach to measuring the spike in volatility. We looked at the max drawdown in the S&P 500 over the 30 days from the beginning and end of each sub-15 period. As the table below shows, the S&P has experienced an average drawdown of about 2.8% after a sub-15 period ends. As previously mentioned, the most recent sub-15 period happened in late August/early September, from the start of that period on the S&P experienced a max drawdown of 2.32%. From a percentage standpoint, the change in the S&P is not nearly as large as seen in the VIX, that being said, a drawdown of 2.8% is more than we expect to see in a typical 30-day period.
